Age-related differences in fall migration timing and performance of juvenile and adult Wood Thrushes departing from a breeding site

Juvenile passerines are expected to have lower migration performance than adults due to their inexperience with long-distance flights and morphological limitations, such as shorter wing length. From 2016-2019 we radio-tagged nestling and adult Wood Thrushes (Hylocichla mustelina) at a breeding site in southwestern Ontario and used the automated Motus Wildlife Tracking System to test if age class predicts timing of the onset of fall migration (date, time of night), flight speed during the initial migration flight across Lake Erie, and overall pace of migration southward through the eastern United States. We detected 60/117 (51%) adults and 82/119 (69%) juveniles departing the breeding area as they initiated fall migration. Compared with adults, juveniles departed significantly earlier in fall and significantly later in the evening. When crossing Lake Erie on their first migration flight juveniles travelled about 25% slower than adults but this was due primarily to adults making better use of tailwinds. When travelling south through the eastern U.S. juveniles had a significantly slower overall migration pace (47.3 ± 5.1km/day) than adults (71.6 ± 4.7km/day). Although we found some evidence that juvenile Wood Thrushes have lower fall migration performance than adults, identifying the proximate and ultimate mechanisms remains a challenge. There is no evidence that juvenile Wood Thrushes are inefficient in migration flight or refueling at stopovers, and it is unlikely that the fall migration pace in this species affects their ability to compete for wintering food resources. More tracking studies from breeding sites are needed to understand the ecological factors favouring and biological significance of, age-related differences in migration performance.

Methods
Radio-tagging

When nestlings were 10 days old, shortly before fledging which occurs at 12-14 days, they were removed from the nest and banded with both Canadian Wildlife Service numbered aluminum bands and unique color band combinations. The largest nestling (by mass) had a blood sample (<25 uL) taken for genetic sexing (HealthGene Molecular Diagnostic and Research Center, Concord, Canada) and was equipped with a uniquely coded radio transmitter using a backpack leg loop harness (Rappole and Tipton 1991) made from 2.5mm Teflon tubing. Only one nestling was tagged at most nests (82%; 131 of 160 nests) but two nestlings were tagged at 29 nests that fledged late in the season. Of all tagged nestlings (n = 189), 133 survived to 16 days post-fledging and 119 survived to the onset of fall migration. We refer to this age class as juvenile.

Adults were caught by placing mist nets in the vicinity of nests and then banded with both Canadian Wildlife Service numbered aluminum bands and unique color band combinations. As with the nestlings, breeding adults (n = 117; 32 male and 85 female) were equipped with a uniquely coded radio transmitter using a backpack leg loop harness (Rappole and Tipton 1991) made from 2.5mm Teflon tubing. Sex was determined by checking for the presence of a brood patch or cloacal protuberance and age was determined as either second year or after second year using plumage criteria (Pyle 1997).

The specifications of the radio-tags varied among years (2016: NTQBW-6-1 (1.6g); 2017: ANTC-M6-1 (1.7g);  2018: ANTC-M6-1 and NTQB2-6-1 (1.6g); 2019: NTQB2-4-2S (1.5g)) due to the manufacturer (Lotek Wireless Inc., Newmarket, Canada) discontinuing models, but all had similar performance, a burst rate of 12.7 seconds, and minimum expected lifespan of 400 days.

Motus Migration Detections

Motus receivers sometimes record false detections due to random radio noise, duplicate tags, and overlapping tag signals when multiple tags are transmitting in the same area (Crewe et al. 2018). A number of filtering and quality control steps were taken to identify these and exclude them from analysis. First, we eliminated detections of fewer than 3 consecutive tag bursts because they are likely to be false detections (Crewe et al. 2018). Next, we eliminated any detection that occurred in impossible locations based on prior knowledge of migration timing and routes from geolocator studies (Stutchbury et al. 2011, Stanley et al. 2015).

 Detections representing the initiation of fall migration were those that occurred after sunset during the migratory period (August 25 to October 15), when a bird was not detected in Ontario again until the following year. We used the time stamp of the first detection on the migration date to represent the time of day of the start of migration. Because the timing of sunset (the moment the sun disappears below the horizon) varies greatly during the migratory period, we calculated the number of minutes after sunset each bird began its migration at and used this number as our response variable. Motus towers cannot determine when a bird actually begins a migratory flight, only when it is first detected. For this reason, we removed one extreme outlier from analysis, an adult that was detected 84 minutes later than the next latest bird, more than 5 hours after sunset.

Our study area lies on the north shore of Lake Erie which gave us the unique opportunity to estimate flight speed of the initial migratory flight of birds that were detected departing the study area and subsequently detected along the south shore of Lake Erie. We cannot assume that the movements between stations are linear, but we can be reasonably certain that birds did not land in the water between detection at these towers. Signal strength readings when a bird flies past a tower create a parabolic arc with increasing strength heading toward the tower and decreasing strength after passing the tower. Because there is variation in the range at which a tower can detect a bird, we calculated the amount of time it took a bird to fly from peak signal strength at the last tower on the north shore of Lake Erie to the peak signal strength at the first tower on the south shore of Lake Erie (Begin-Marchand et al. 2021; Additional Figure 1.). We measured the distance between the two towers and using the time elapsed between peak detections, calculated the average flight speed (km/h) and used this as our response variable. Some birds were detected before and after crossing Lake Erie, but were not picked up by one or both towers long enough to show the expected rising and falling of signal strength. These individuals (n=11 adults; 8 juveniles) were removed from analysis because without a clear peak in signal strength, we could not accurately estimate the time when they passed each tower.

The direction and speed of wind was found to be an important predictor of migratory flight speed (Mitchell et al. 2015) so we included it as a covariate in our flight speed models. Following the methods of Morbey et al. 2018, we used the RNCEP package (Kemp et al. 2012) in R to extract easterly (90°) and northerly (0°) wind components at the 925mb pressure level, which corresponds to an altitude of 675­–825m over Lake Erie and is comparable to the altitude of migratory flights measured in other species of thrushes (Bowlin et al. 2015). We used linear interpolation with the NCEP.interpol function to interpolate 925mb wind conditions at the time and date of departure for each bird. To estimate the tailwind component, we used Vw × cos(β), where Vw is the wind speed and β is the difference between the flight bearing and wind direction (Safi et al. 2013). We assumed a flight bearing of 180°, as most birds were detected almost directly south of the study area after crossing the lake. This resulted in a tailwind metric that ranged from -10 when flying directly into a strong headwind, to 10 when benefitting from a strong and direct tailwind.

The last measure we examined was the pace of migration through the U.S. after the night of the initial migratory flight. Motus does not allow us to determine when a bird is stopping over (e.g., relatively stationary for one or more nights) versus migrating because of gaps in tower coverage, but we can measure the time it takes a bird to travel between towers and measure the distance between those towers to calculate the average number of kilometers a bird travels per day. If a bird is spending more time stopping over, it will travel fewer kilometers per day when covering the distance between towers. To ensure that we were measuring stopover behavior and not speed of direct overnight flights, we eliminated detections that occurred on consecutive days (n=2). We detected 25 individuals at multiple towers, not on successive days, as they migrated south through the United States. In Swainson’s Thrushes (Catharus ustulatus) migration pace was found to be slower at higher latitudes (Begin-Marchand et al. 2021), so we calculated the midpoint in latitude between Motus towers to include as a variable in analysis. We also included the distance between Motus towers because it was also found to be important to predicting migration pace in Swainson’s Thrushes, with a higher pace when detections were geographically closer to each other (Begin-Marchand et al. 2021).